Hear Seth Clark explain how the largest Military, Civil government, and Enterprises design, build, and deploy large-scale AI projects from the lab into production in a way that is quick, safe, and secure. Topics include: the AI Pipeline; the differences between model management, Model Ops, and MLOps; what all the members of an AI team should do (and more importantly, NOT do); how the answer is not build-or-buy, but build-AND-buy; what new threats exist in an AI application and how to defend against them; as well as how to build AI systems that can explain their decisions to humans.
Watch on Youtube
Listen on Simplecast•
- 00:00 Introduction
- 00:49 The genesis story of Modzy – born from Booz Allen consulting engagements
- 03:24 Working with Military, civil government, finance, oil & gas, energy and utilities
- 05:28 Why do customers need a platform? Why not just DIY?
- 08:30 Description of Modzy – the platform and model marketplace
- 15:00 The AI pipeline – from big idea to data collection, model development, training, deployment, assessment, retraining, explainability
- 21:00 What is the difference between model management, Model Ops, and MLOps?
- 22:42 Deploy to where? Tactical edge, public cloud, private datacenter, air-gapped datacenter
- 25:30 AI is a team sport – data scientists, software developers, machine learning engineers, business analysts, executives
- 30:48 Model Marketplace – an app store for AI models
- 33:42 AI requires adversarial defense mechanisms to protect against new and different attacks
- 37:03 What is explainable AI, why do we need it, and how do you achieve it?
- 41:45 Ideal customer for a platform like this
- 44:49 Wrap-up